From mboxrd@z Thu Jan 1 00:00:00 1970 Received: from eggs.gnu.org ([2001:4830:134:3::10]:60104) by lists.gnu.org with esmtp (Exim 4.71) (envelope-from ) id 1bUrpp-0000eJ-8B for qemu-devel@nongnu.org; Wed, 03 Aug 2016 04:47:30 -0400 Received: from Debian-exim by eggs.gnu.org with spam-scanned (Exim 4.71) (envelope-from ) id 1bUrpm-0003Fq-2Q for qemu-devel@nongnu.org; Wed, 03 Aug 2016 04:47:29 -0400 Received: from mx1.redhat.com ([209.132.183.28]:47344) by eggs.gnu.org with esmtp (Exim 4.71) (envelope-from ) id 1bUrpl-0003Fd-Q7 for qemu-devel@nongnu.org; Wed, 03 Aug 2016 04:47:25 -0400 Date: Wed, 3 Aug 2016 10:47:21 +0200 From: Igor Mammedov Message-ID: <20160803104721.2e6d453c@nial.brq.redhat.com> In-Reply-To: <49831bf2-565c-88ef-9ab2-ce6bf7974ac7@redhat.com> References: <1470211497-116801-1-git-send-email-imammedo@redhat.com> <49831bf2-565c-88ef-9ab2-ce6bf7974ac7@redhat.com> MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: quoted-printable Subject: Re: [Qemu-devel] [PATCH] pc: pcihp: avoid adding ACPI_PCIHP_PROP_BSEL twice List-Id: List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , To: Marcel Apfelbaum Cc: qemu-devel@nongnu.org, marcandre.lureau@redhat.com, "Michael S. Tsirkin" , Paolo Bonzini , Richard Henderson , Eduardo Habkost On Wed, 3 Aug 2016 11:17:11 +0300 Marcel Apfelbaum wrote: > On 08/03/2016 11:04 AM, Igor Mammedov wrote: > > PCI hotplug for bridges was introduced only since 2.0 however > > acpi_set_bsel()->object_property_add_uint32_ptr(bus, ACPI_PCIHP_PROP_= BSEL) > > didn't take in account that for legacy mode (1.7) when > > PCI hotplug for bridges is unavailable and ACPI_PCIHP_PROP_BSEL property > > the only bus "PCI.0' has been created earlier at acpi_pcihp_init() time. > > > > We managed to live with it only because of error rised by adding > > a duplicate property in acpi_set_bsel() has been ignored which > > resulted in useless leaking of just allocated (int)bus_bsel. > > > > Issue affects only 1.7 machine type as ACPI tables supported by > > QEMU were introduced at that time, but there wasn't PCI hotplug > > for bridges till the next release (2.0). > > > > Fix it by removing duplicate ACPI_PCIHP_PROP_BSEL intialization > > in acpi_pcihp_init() and doing it only in one place acpi_set_pci_info(). > > > > PS: > > do not ignore error returned by object_property_add_uint32_ptr() > > and abort QEMU since it's programming error which should be fixed > > instead of being ignored. > > > > Signed-off-by: Igor Mammedov > > Reported-by: Marc-Andr=C3=A9 Lureau > > --- > > hw/acpi/pcihp.c | 10 ---------- > > hw/i386/acpi-build.c | 4 ++-- > > 2 files changed, 2 insertions(+), 12 deletions(-) > > > > diff --git a/hw/acpi/pcihp.c b/hw/acpi/pcihp.c > > index d957d1e..3298d94 100644 > > --- a/hw/acpi/pcihp.c > > +++ b/hw/acpi/pcihp.c > > @@ -302,16 +302,6 @@ void acpi_pcihp_init(Object *owner, AcpiPciHpState= *s, PCIBus *root_bus, > > s->root=3D root_bus; > > s->legacy_piix =3D !bridges_enabled; > > > > - if (s->legacy_piix) { > > - unsigned *bus_bsel =3D g_malloc(sizeof *bus_bsel); > > - > > - s->io_len =3D ACPI_PCIHP_LEGACY_SIZE; > > - > > - *bus_bsel =3D ACPI_PCIHP_BSEL_DEFAULT; > > - object_property_add_uint32_ptr(OBJECT(root_bus), ACPI_PCIHP_PR= OP_BSEL, > > - bus_bsel, NULL); > > - } > > - > > memory_region_init_io(&s->io, owner, &acpi_pcihp_io_ops, s, > > "acpi-pci-hotplug", s->io_len); > > memory_region_add_subregion(address_space_io, s->io_base, &s->io); > > diff --git a/hw/i386/acpi-build.c b/hw/i386/acpi-build.c > > index a26a4bb..7174933 100644 > > --- a/hw/i386/acpi-build.c > > +++ b/hw/i386/acpi-build.c > > @@ -438,7 +438,7 @@ static void *acpi_set_bsel(PCIBus *bus, void *opaqu= e) > > > > *bus_bsel =3D (*bsel_alloc)++; > > object_property_add_uint32_ptr(OBJECT(bus), ACPI_PCIHP_PROP_BS= EL, > > - bus_bsel, NULL); > > + bus_bsel, &error_abort); =20 >=20 > Hi Igor, >=20 > I think the above hunk should be in a separate patch, but > since the patch itself is small enough I am OK with it. I've been conflicted whether to split or not but decided not to as it exposes issue that the patch fixes and splitting could lead to bisection issues in case this part comes first. >=20 >=20 > > } > > > > return bsel_alloc; > > @@ -447,7 +447,7 @@ static void *acpi_set_bsel(PCIBus *bus, void *opaqu= e) > > static void acpi_set_pci_info(void) > > { > > PCIBus *bus =3D find_i440fx(); /* TODO: Q35 support */ > > - unsigned bsel_alloc =3D 0; > > + unsigned bsel_alloc =3D ACPI_PCIHP_BSEL_DEFAULT; > > > > if (bus) { > > /* Scan all PCI buses. Set property to enable acpi based hotpl= ug. */ > > =20 >=20 >=20 >=20 > Reviewed-by: Marcel Apfelbaum Thanks! >=20 > Thanks, > Marcel